Analyses and figures used in the paper: Globally consistent reef size spectra with fishes and invertebrates

When using these analyses, please cite: Heather et al. (2020) Globally consistent reef size spectra with fishes and invertebrates

Main text

Conceptual diagram

# Conceptual diagram of the hypotheses. Made using simulated data

set.seed(1)

# setting parameters for normal distribution plots (for steepening effect)
invr_mean <- rlnorm(n = 10, meanlog = 1.8, sdlog = 0.5) + 10
vert_mean <- rlnorm(n = 10, meanlog = 2, sdlog = 1) + 15
invr_sd   <- rnorm(n = 10, mean = 8, sd = 1)
vert_sd   <- rnorm(n = 10, mean = 10, sd = 1)

# Figure 1A
p1 <-
  tibble(type = "invr", spp =1:10, mean = invr_mean, sd = invr_sd) %>% 
  bind_rows(tibble(type = "vert", spp =1:10, mean = vert_mean, sd = vert_sd)) %>% 
  mutate(vals = map2(mean, sd, .f=~dnorm(0:100, mean=.x, sd=.y))) %>% 
  unnest(cols = "vals") %>% 
  mutate(id = paste0(type, spp)) %>% 
  mutate(x = rep(0:100, 20)) %>% 
  ggplot(aes(x, vals, fill=type, group=id)) +
  geom_area(position = 'identity', alpha=0.2, aes(col=type)) +
  scale_x_continuous(expand = c(0, 0)) +
  scale_y_continuous(breaks = NULL, expand = c(0, 0)) +
  theme_classic(24) +
  theme(panel.grid = element_blank(),
        panel.border = element_blank(), 
        axis.text = element_blank(),
        axis.ticks = element_blank(), 
        legend.position = c(0.8, 0.8), legend.background = element_rect(fill = "transparent"),
        axis.line = element_line(arrow = arrow(type = "open"))) +
  labs(x = "Log(mass)",
       y = "Probability density") +
  scale_fill_npg(name="Species", labels = c("Invertebrate", "Fish"), 
                 guide = guide_legend(override.aes = list(alpha = 1))) + 
  scale_colour_npg(guide = "none")

# Figure 1B
p2 <-
  tibble(mass = 1:10, invr = (8 + (-1.2*mass)), vert =  (6 + (-1*mass))) %>% 
  gather(key = "type", value = "abundance", -mass) %>% 
  ggplot(aes(mass, abundance, col=type)) +
  geom_ribbon(aes(ymin=(6 + (-1*mass)), ymax=(8 + (-1.2*mass))), alpha=0.3, col="transparent", fill = mypal[1]) +
  geom_line(size=3) + 
  geom_point(aes(fill = type), col="black", shape=21, size=3) + 
  theme_classic(24) +
  theme(panel.grid = element_blank(),
        panel.border = element_blank(), 
        axis.text = element_blank(),
        axis.ticks = element_blank(), 
        axis.line = element_line(arrow = arrow(type = "open")), 
        legend.position = c(0.3, 0.2), 
        legend.background = element_rect(fill = "transparent")) + 
  labs(x = "Log(mass)",
       y = "Log(abundance)") +
  scale_colour_npg(name="Community", labels = c("Fish and Invertebrates", "Fish-only")) + 
  scale_fill_npg(guide = "none")

# set new distribution params (for shallowing effect)
invr_sd <- rnorm(n = 10, mean = 8, sd = 1.3)
invr_mean <- rlnorm(n = 10, meanlog = 1.8, sdlog = 0.8) + 20

# Figure 1C
p3 <- 
  tibble(type = "invr", spp =1:10, mean = invr_mean, sd = invr_sd) %>% 
  bind_rows(tibble(type = "vert", spp =1:10, mean = vert_mean, sd = vert_sd)) %>% 
  mutate(vals = map2(mean, sd, .f=~dnorm(0:100, mean=.x, sd=.y))) %>% 
  unnest(cols = "vals") %>% 
  mutate(id = paste0(type, spp)) %>% 
  mutate(x = rep(0:100, 20)) %>% 
  ggplot(aes(x, vals, fill=type, group=id)) +
  geom_area(position = 'identity', alpha=0.2, aes(col=type)) +
  scale_x_continuous(expand = c(0, 0)) +
  scale_y_continuous(breaks = NULL, expand = c(0, 0)) +
  theme_classic(24) +
  theme(panel.grid = element_blank(),
        panel.border = element_blank(), 
        axis.text = element_blank(),
        axis.ticks = element_blank(), 
        legend.position = c(0.8, 0.8), 
        legend.background = element_rect(fill = "transparent"),
        axis.line = element_line(arrow = arrow(type = "open"))) +
  labs(x = "Log(mass)",
       y = "Probability density") +
  scale_fill_npg(name="Species", labels = c("Invertebrate", "Fish"), 
                 guide = guide_legend(override.aes = list(alpha = 1))) + 
  scale_colour_npg(guide = "none")

# Figure 1D
p4 <-
  tibble(mass = 1:10, invr = (6.2 + (-0.8*mass)), vert =  (6 + (-1*mass))) %>% 
  gather(key = "type", value = "abundance", -mass) %>% 
  ggplot(aes(mass, abundance, col=type)) +
  geom_ribbon(aes(ymin=(6 + (-1*mass)), ymax=(6.2 + (-0.8*mass))), alpha=0.3, col="transparent", fill = mypal[1]) +
  geom_line(size=3) + 
  geom_point(aes(fill = type), col="black", shape=21, size=3) + 
  theme_classic(24) +
  theme(panel.grid = element_blank(),
        panel.border = element_blank(), 
        axis.text = element_blank(),
        axis.ticks = element_blank(), 
        axis.line = element_line(arrow = arrow(type = "open")), 
        legend.position = c(0.3, 0.2), 
        legend.background = element_rect(fill = "transparent")) + 
  labs(x = "Log(mass)",
       y = "Log(abundance)") +
  scale_colour_npg(name="Community", labels = c("Fish and Invertebrates", "Fish-only")) + 
  scale_fill_npg(guide = "none")

# Combined figure
p5 <- plot_grid(p1, p2, p3, p4, labels = c('A', 'B','C', 'D'), label_size = 24, ncol = 2)

# Adding the annotations (curves, text, images)
fig_1 <-
  ggdraw(p5) +
  geom_curve(aes(xend = 0.59, yend = 0.9, x = 0.175, y = 0.89), 
             size=4, 
             col ="black", 
             arrow = arrow(length = unit(0.03, "npc")), 
             lineend="round", 
             curvature = -0.2) +
  geom_curve(aes(xend = 0.59, yend = 0.9, x = 0.175, y = 0.89), 
             size=3, 
             col = mypal[1], 
             arrow = arrow(length = unit(0.03, "npc")), 
             lineend="round", 
             curvature = -0.2) +
  geom_curve(aes(xend = 0.85, yend = 0.225, x = 0.25, y = 0.425), 
             size=4, col ="black", 
             arrow = arrow(length = unit(0.03, "npc")), 
             lineend="round", 
             curvature = -0.4) +
  geom_curve(aes(xend = 0.85, yend = 0.225, x = 0.25, y = 0.425), 
             size=3, col = mypal[1],
             arrow = arrow(length = unit(0.03, "npc")), 
             lineend="round", 
             curvature = -0.4) +
  draw_image("fish_silouhette_plotblue.png", x = 0.4, y = 0.75, 
             hjust = 1, vjust = 1, width = 0.08, height = 0.08) +
  draw_image("invert_silouhette_plotred.png", x = 0.1, y = 0.95, 
             hjust = 1, vjust = 1, width = 0.04, height = 0.04) +
  draw_image("fish_silouhette_plotblue.png", x = 0.4, y = 0.25, 
             hjust = 1, vjust = 1, width = 0.08, height = 0.08) +
  draw_image("invert_silouhette_plotred.png", x = 0.25, y = 0.5,
             hjust = 1, vjust = 1, width = 0.06, height = 0.06) 

# save figure and include in markdown
save_plot(filename = "output/figs/fig1.png", plot = fig_1, dpi=300, base_height = 10)
knitr::include_graphics("output/figs/fig1.png")

Figure 1: Hypothesized effect of including invertebrates in the size spectrum: 1) A steepening effect (A, B), and 2) a shallowing effect (C, D). The steepness of the size spectrum arises from the relative abundances of larger and smaller bodied individuals. If invertebrates have a shallower size spectrum slope (i.e. relatively fewer large-bodied individuals) compared to their co-located fish (A), we would expect the slope of the size spectrum of the combined community (fish and invertebrates) to be steeper than the slope of the fish only (B). A shallowing effect (D) would be expected if invertebrates have a relatively greater number of large-bodied individuals compared to the fish-only community (C).

Normalized size spectrum

# Models
nass_mod_v <- read_rds("data/nass_mod_v.rds")
nass_mod_c <- read_rds("data/nass_mod_c.rds")

# params of models
vert_slope <- nass_mod_v %>% summary() %>% coef() %>% .[2, "Estimate"]
vert_incpt <- nass_mod_v %>% summary() %>% coef() %>% .[1, "Estimate"]
comb_slope <- nass_mod_c %>% summary() %>% coef() %>% .[2, "Estimate"]
comb_incpt <- nass_mod_c %>% summary() %>% coef() %>% .[1, "Estimate"]

draw_key_cust <- function(data, params, size) {
  data$fill <- data$colour
  draw_key_rect(data, params, size)
}

p7 <- 
  fread("data/nass.csv") %>% 
  ggplot(aes(bin_mid, norm_density, colour = group_ordered)) + 
  geom_point(alpha = 0.05, size = 3, key_glyph = draw_key_cust) +
  geom_abline(slope = vert_slope, intercept = vert_incpt, alpha  = 1, lwd = 2, colour = mypal[2]) +
  geom_abline(slope = comb_slope, intercept = comb_incpt, alpha  = 1, lwd = 2, colour = mypal[1]) +
  xlab("Mass bin (g)") +
  ylab(expression(paste("Abundance (", m^-2, ")"))) +
  scale_colour_manual(name = "Community",
                      labels = c("Fish-only", "Fish and invertebrates"),
                      values = c(mypal[2], mypal[1])) +
  theme_bw(40) +
  annotation_logticks() +
  scale_x_continuous(breaks = trans_breaks("log2", function(x) 2^x), 
                     trans = "log2", 
                     labels = trans_format("log2", math_format(2^.x))) +
  scale_y_continuous(breaks = trans_breaks("log2", function(x) 2^x), 
                     trans = "log2", 
                     labels = trans_format("log2", math_format(2^.x))) +
  guides(colour = guide_legend(override.aes = list(alpha = 1))) +
  theme(
    axis.title.x = element_text(hjust = 0.45, vjust=0),
    legend.position = "none")

fig_2 <- 
  ggdraw(p7) +
  draw_label("Community", colour = "black", size = 30, x = 0.825, y = 0.825) +
  draw_image("fish_silouhette_plotblue.png", x = 0.8, y = 0.8, hjust = 1, vjust = 1, width = 0.08, height = 0.08) +
  draw_image("invert_silouhette_plotred.png", x = 0.9, y = 0.75, hjust = 1, vjust = 1, width = 0.06, height = 0.06) +
  draw_label("+", colour = mypal[1], size = 30, x = 0.825, y = 0.715) +
  draw_image("fish_silouhette_plotred.png", x = 0.8, y = 0.75, hjust = 1, vjust = 1, width = 0.08, height = 0.08)


save_plot(filename = "output/figs/fig_2.png", plot = fig_2, base_height = 10, dpi=300)
knitr::include_graphics("output/figs/fig_2.png")

Figure 2: Invertebrates steepen the normalized abundance size spectrum. Separate normalized abundance size spectra are shown for the fish-only and combined (fish and invertebrate) communities, with solid lines representing fits from linear mixed effects models for the global data (“Site” nested within “Ecoregion” as random effects). Fish-only slope = \(-1.73 \pm 0.06\), combined slope = \(-1.92 \pm 0.04\). Points have been offset on the x-axis for clarity.

Map

# get the natural earth data
load("data/NaturalEarth.RData")

# set the projection type to 'Robinson'
PROJ <- "+proj=robin +lon_0=0 +x_0=0 +y_0=0 +ellps=WGS84 +datum=WGS84 +units=m +no_defs" 
NE_countries_rob  <- spTransform(NE_countries, CRSobj = PROJ)
NE_graticules_rob <- spTransform(NE_graticules, CRSobj = PROJ)
NE_box_rob        <- spTransform(NE_box, CRSobj = PROJ)

# project latitudes and longitudes
prj.coord <- project(cbind(lbl.Y$lon, lbl.Y$lat), proj=PROJ)
lbl.Y.prj <- cbind(prj.coord, lbl.Y)
names(lbl.Y.prj)[1:2] <- c("X.prj","Y.prj")

# add in the slope data
df.ss <- fread("data/slope_diff.csv") 
prj.coord.ss <- cbind(df.ss, project(cbind(df.ss$lon, df.ss$lat), proj=PROJ)) %>% filter(!is.na(slope_diff))

# # getting summary statistics 
floor_dec   <- function(x, level=1) round(x - 5*10^(-level-1), level)
ceiling_dec <- function(x, level=1) round(x + 5*10^(-level-1), level)
slope_stats <- df.ss %>% pull(slope_diff) %>% summary()
slope_min   <- slope_stats["Min."] %>% floor_dec()
slope_max   <- slope_stats["Max."] %>% ceiling_dec()

# Main map fig
fig_3B <- 
  ggplot() +
  geom_polygon(data=NE_box_rob, aes(x=long, y=lat), colour="gray80", fill="transparent", size = 0.25) +
  geom_path(data=NE_graticules_rob, aes(long, lat, group=group), linetype="dotted", color="grey50", size = 0.25) +
  geom_polygon(data=NE_countries_rob, aes(long,lat, group=group), colour="gray80", fill="gray80", size = 0.25) +
  coord_fixed(ratio = 1) +
  theme_void() +
  geom_point(data = prj.coord.ss, aes(x=V1, y=V2, colour=slope_diff), size=4, alpha=1) +
  theme(legend.position = 'bottom', 
        legend.spacing.x = unit(0.5, 'cm'),
        legend.text = element_text(margin = margin(t = 10)))+
  scale_colour_continuous(limits = c(slope_min, slope_max),
                          breaks = seq(slope_min,slope_max, by = 0.1)) +
  scale_colour_viridis_c(option = "magma") +
  guides(col = guide_colourbar(title = expression(paste("Invertebrate inclusion effect (", Delta, lambda, ")")),
                               label.position = "bottom",
                               title.position = "top",
                               title.vjust = 1,
                               title.hjust = 1, 
                               frame.colour = "black", 
                               ticks.colour = "black",
                               ticks.linewidth = 2,
                               barwidth = 30,
                               barheight = 1.5)) +
  theme(text = element_text(size=30)) +
  theme(plot.margin=grid::unit(c(0,0,0,0), "mm"))

# Slope difference barplot
fig_3C <-  
  fread("data/lat_bins_slopes.csv") %>% 
  ggplot(aes(x = lat_bin,  y = mean_slope_diff, fill = mean_slope_diff)) +
  geom_bar(stat = "identity") + 
  geom_errorbar(aes(ymin = mean_slope_diff - sd_slope_diff, 
                    ymax = mean_slope_diff + sd_slope_diff), 
                width=0, colour="black", alpha=0.9, size=0.8) +
  scale_fill_viridis_c(option = "magma", limits = c(slope_min, slope_max)) +
  coord_flip() +
  theme_classic(20) +
  theme(axis.title.y=element_blank(),
        axis.text.y=element_blank(),
        axis.ticks.y=element_blank(),
        axis.line.x.bottom = element_line(color = 'black'),
        axis.line.y.left   = element_blank(),
        panel.background = element_rect(fill = "transparent"), # bg of the panel
        plot.background = element_rect(fill = "transparent", color = NA), # bg of the plot
        panel.grid.major = element_blank(), # get rid of major grid
        panel.grid.minor = element_blank(), # get rid of minor grid
        legend.background = element_rect(fill = "transparent"), # get rid of legend bg
        legend.box.background = element_rect(fill = "transparent"), # get rid of legend panel bg
        legend.position = "none",
        aspect.ratio = 2,
        axis.title.x = element_text(size = 20, colour = "transparent", margin=margin(b=20)))



# Slope fish-only vs combined barplot
fig_3A <- 
  fread("data/lat_bins_slopes.csv") %>% 
  ggplot(aes(y = mean_slope_comb, x=lat_bin)) + 
  geom_bar(stat = "identity", aes(fill = mypal[1]), col = "black") +
  geom_bar(aes(y = mean_slope_vert, x = lat_bin, fill = mypal[2]), stat="identity", col = "black") +
  coord_flip() +
  theme_classic(20) +
  theme(axis.title.y=element_blank(),
        axis.text.y=element_blank(),
        axis.ticks.y=element_blank(),
        axis.line.x.bottom = element_line(color = 'black'),
        axis.line.y.left   = element_blank(),
        panel.background = element_rect(fill = "transparent"), # bg of the panel
        plot.background = element_rect(fill = "transparent", color = NA), # bg of the plot
        panel.grid.major = element_blank(), # get rid of major grid
        panel.grid.minor = element_blank(), # get rid of legend panel bg
        legend.position = "none") +
  ylab(expression(paste("Mean slope (", bar(lambda), ")"))) +
  theme(aspect.ratio = 2) +
  scale_fill_identity()

# adding annotations (images)
fig_3A_edit <- 
  ggdraw(fig_3A) + 
  draw_image("fish_silouhette_plotblue.png", x = 0.63, y = 1.03, hjust = 1, vjust = 1, width = 0.1, height = 0.1) +
  draw_image("invert_silouhette_plotred.png", x = 0.49, y = 1.02, hjust = 1, vjust = 1, width = 0.07, height = 0.07) +
  draw_label("+", colour = mypal[1], size = 30, x = 0.4, y = 0.975)

# adding axis title 
fig_3C_edit <-
  ggdraw(fig_3C) +
  draw_label("Mean invertebrate", x = 0.5, y = 0.065, size = 20) + # use relative coordinates for positioning
  draw_label(expression(paste("inclusion effect (", bar(Delta), bar(lambda), ")")), x = 0.5, y = 0.015, size = 20)

# making the edits into grobs to be added to main fig
fig_3A_grob = ggplotGrob(fig_3A_edit)
fig_3C_grob = ggplotGrob(fig_3C_edit)

# useful numbers when adding annotations
min_lon <- project(cbind(-180,-90), proj=PROJ) %>% .[,1]
min_lat <- project(cbind(-180,-90), proj=PROJ) %>% .[,2]
max_lon <- project(cbind(180,90), proj=PROJ) %>% .[,1]
max_lat <- project(cbind(180,90), proj=PROJ) %>% .[,2]

# final figure all combined
fig_3 <-
  fig_3B + 
  xlim(c(-2.2e7, 2.2e7)) + 
  annotation_custom(grob = fig_3A_grob, 
                    xmin = min_lon - 20300000, 
                    xmax = max_lon - 20300000, 
                    ymin = min_lat - 200000,
                    ymax = max_lat - 1500000) + 
  annotation_custom(grob = fig_3C_grob, 
                    xmin = min_lon - 20300000 + 4e7, 
                    xmax = max_lon - 20300000 + 4e7, 
                    ymin = min_lat - 300000,
                    ymax = max_lat - 1500000) +
  annotation_custom(text_grob("A", size = 30),  
                    xmin = min_lon - 20300000, 
                    xmax = max_lon - 20300000 - 5e6, 
                    ymin = min_lat - 200000,
                    ymax = max_lat - 2200000 + 1.8e7) +
  annotation_custom(text_grob("B", size = 30),  
                    xmin = min_lon - 20300000, 
                    xmax = max_lon - 20300000 + 1e7, 
                    ymin = min_lat - 200000,
                    ymax = max_lat - 2200000 + 1.8e7) +
  annotation_custom(text_grob("C", size = 30),  
                    xmin = min_lon - 20300000, 
                    xmax = max_lon - 20300000 + 7.5e7, 
                    ymin = min_lat - 200000,
                    ymax = max_lat - 2200000 + 1.8e7) 
  

save_plot(filename = "output/figs/fig_3.png", plot=fig_3, base_height = 10, dpi=300)
knitr::include_graphics("output/figs/fig_3.png")

Figure 3: The inclusion of invertebrates results in a consistent community size spectrum slope of \(\sim -2\). (A) The size spectrum slope for fish-only communities (blue) and when including invertebrates (orange). (B) A map of the invertebrate inclusion effect (\(\Delta \lambda\)) across the globe. (C) The ‘invertebrate inclusion effect’ (\(\Delta \lambda\)) on the size spectrum slope varies with latitude, steeper spectra when including invertebrates at high latitudes where more negative represents a greater steepening when including invertebrates in the community size spectrum.. Each bar in A and C represents the mean over 5 degrees of latitude. Error bars in (C) represent the 95% confidence intervals; missing error bars represent insufficient data.

Spatial scales

var_outputs <- fread("data/var_outputs.csv")

line_comb_y1 <- var_outputs %>% filter(scale == "Ecoregion", community == "Combined") %>% pull(slope) %>% max()
line_comb_y2 <- var_outputs %>% filter(scale == "Ecoregion", community == "Combined") %>% pull(slope) %>% min()
line_comb_y3 <- var_outputs %>% filter(scale == "Site", community == "Combined") %>% pull(slope) %>% median()

line_vert_y1 <- var_outputs %>% filter(scale == "Ecoregion", community == "Vertebrate") %>% pull(slope) %>% max()
line_vert_y2 <- var_outputs %>% filter(scale == "Ecoregion", community == "Vertebrate") %>% pull(slope) %>% min()
line_vert_y3 <- var_outputs %>% filter(scale == "Site", community == "Vertebrate") %>% pull(slope) %>% median()

fig_4 <- 
  var_outputs %>% 
  ggplot(aes(x = x, y = slope, fill = community, group = fact)) +
  geom_hline(yintercept = -2, lty = 2) +
  geom_violin() +
  geom_boxplot(width = 0.1, data = var_outputs %>% filter(scale != "Global"), key_glyph = draw_key_rect) +
  geom_point(data = var_outputs %>% filter(scale == "Global"), 
             colour="black", 
             shape=21, 
             size = 5, 
             key_glyph = draw_key_rect) +
  theme_bw(40) +
  labs(y = expression(lambda),
       x = element_blank()) +
  scale_x_continuous(breaks = c(2,4,6), 
                     labels = c("Global\n(overall mean)", "Ecoregion\n(across Ecoregion)", "Site\n(within Ecoregion)")) +
  scale_fill_manual(values = c(mypal[2], mypal[1]), name = "Community", labels = c("Fish-only", "Fish and invertebrates")) +
  theme(
    axis.title.y = element_text(angle = 0, vjust = 0.5),
    axis.title.x = element_text(hjust = 0.45, vjust=0),
    legend.position = c(0.18, 0.85), 
    legend.background = element_rect(fill = "transparent"), 
    legend.key.size = unit(1, "cm"), 
    legend.key = element_rect(colour = "black", fill = NA),
    legend.title = element_text(size = 40),
    legend.text = element_text(size = 25)) +
  annotate("segment", x = 5.5, y = line_vert_y3, xend = 3.5, yend = line_vert_y1, lty = 2, alpha=0.4) +
  annotate("segment",x = 5.5, y = line_vert_y3, xend = 3.5, yend = line_vert_y2, lty = 2, alpha=0.4) +
  annotate("segment",x = 6.5, y = line_comb_y3, xend = 4.5, yend = line_comb_y1, lty = 2, alpha=0.4) +
  annotate("segment",x = 6.5, y = line_comb_y3, xend = 4.5, yend = line_comb_y2, lty = 2, alpha=0.4)

save_plot(filename = "output/figs/fig_4.png", plot=fig_4, base_height = 10, dpi=300)
knitr::include_graphics("output/figs/fig_4.png")

Figure 4: The contribution of spatial scale to abundance size spectra slope estimates. “Ecoregion” refers to the variation among ecoregions globally in the linear mixed effects model and “Site” refers to the variation among individual reef sites within ecoregions. Dotted lines between the violins are added to emphasize that the variation at the site level represents the added variation after accounting for the variation at the ecoregion level. A horizontal dotted line at -2 is added to highlight the slope in previous studies based on pelagic studies.

Supplementary material

Mean body size

log(L_{}) = _0 + 1log(L{}) + C +

# Models 
lm_mean  <- lm_mean <- lm(L_mean_log ~ log_lmax + class, data = fread("data/lm_mean.csv")) 

fig_s1_1 <- 
  fread("data/lm_mean.csv") %>% 
  mutate(fit_mean = predict(lm_mean)) %>%
  ggplot(aes(log_lmax, L_mean_log)) +
  geom_point(size = 3) +
  facet_wrap(~class) +
  geom_line(aes(log_lmax, fit_mean), colour = mypal[1], size=2) + 
  labs(x=expression(paste(log(L[infinity]), ", cm")), 
       y=expression(paste(log(L[mu]), ", cm"))) +
  theme_bw(base_size = 30) 

save_plot(filename = "output/figs/suppl_S1_1.png", plot = fig_s1_1, dpi=300, base_height = 10)
knitr::include_graphics("output/figs/suppl_S1_1.png")

Figure S1: The normalized abundance size spectrum for 93 coastal ecoregions. Linear mixed effects models are fitted with site nested within ecoregion as random effects. Each site is individually coloured within the ecoregion. Axes are on the log-log scale.

Standard deviation body size

\(log(L_{\sigma}) = \beta_0 + \beta_1log(L_{\infty})+ \epsilon\)

# Model 
lm_sd <- lm(L_sd_log ~ log_lmax, data = fread("data/lm_sd.csv"))

fig_s1_2 <- 
  fread("data/lm_sd.csv") %>% 
  mutate(fit_mean = predict(lm_sd)) %>%
  ggplot(aes(log_lmax, L_sd_log)) +
  geom_point(size = 3) +
  geom_line(aes(log_lmax, fit_mean), colour=mypal[1], size = 2) + 
  labs(x=expression(paste(log(L[infinity]), ", cm")), 
       y=expression(paste(log(L[sigma]), ", cm"))) +
  theme_bw(base_size = 30) 

save_plot(filename = "output/figs/suppl_S1_2.png", plot = fig_s1_2, dpi=300, base_height = 10)
knitr::include_graphics("output/figs/suppl_S1_2.png")

Figure S2: The relationship between asymptotic size, \(L_\infty\), and the standard deviation of the lognormal distribution, \(L_\sigma\), for 236 invertebrate species across seven classes, with a fitted linear model.

Body size estimates

# Creates a density function for which to plot
create_lnorm_dens <- function(df, meanlog, sdlog){
  data.frame(
    size_class = seq(from = 0, to = max(df$LMAX)*1.5, length = 100)) %>% 
    mutate(prob = dlnorm(size_class, meanlog=unique(meanlog), sdlog=unique(sdlog)))
}

# Probability density based on observed data
obs_dens <- 
  fread("data/size_est.csv") %>% 
  group_by(species_name) %>% 
  nest() %>% 
  mutate(normal_curve = map(data, ~create_lnorm_dens(., .$meanlog, .$sdlog))) %>% 
  select(-data) %>% 
  unnest(cols = c(normal_curve))

# Probability density based on data estimated from LMAX
est_dens <-   
  fread("data/size_est.csv") %>% 
  group_by(species_name) %>% 
  nest() %>% 
  mutate(normal_curve = map(data, ~create_lnorm_dens(., .$meanlog_est, .$sdlog_est))) %>% 
  select(-data) %>% 
  unnest(cols = c(normal_curve))

# All together
fig_s1_3 <-
  fread("data/sizes.csv") %>% 
  ggplot() +
  aes(size_class) +
  geom_histogram(aes(y=..density..), bins = 30, alpha=0.5) +
  facet_wrap(~species_name, scales="free") + 
  geom_line(aes(y = prob), data = obs_dens, colour = mypal[3], size = 2) +
  geom_line(aes(y = prob), data = est_dens, colour = mypal[4], size = 2) +
  geom_vline(aes(xintercept = (LMAX)), col=mypal[5], lty=2, size = 1.5) +
  labs(x = "Size class, cm",
       y = "Probability density") +
  theme_grey(22)

save_plot("output/figs/fig_s1_3.png", fig_s1_3, base_height = 10, dpi=300)
knitr::include_graphics("output/figs/fig_s1_3.png")

Figure S1.3: Estimating invertebrate body size distributions. The observed body length distributions (grey bars) of the 20 invertebrate species with the greatest number of body size estimates. The green line indicates the lognormal distribution fitted to the observed body length data, whilst the blue line represents the estimated lognormal distribution based solely on the asymptotic body length (\(L_\infty\), orange dashed line) of the species and its taxonomic class.

Length-weight conversion

\(log(M_{\infty}) = \beta_0 + \beta_1log(L_{\infty}) + \epsilon\)

\(log(M_{\infty}) = \beta_0 + \beta_1log(L_{\infty}) + C + \epsilon\)

# models
LW_mod1 <- lm(log_mmax ~ log_lmax, fread("data/lw.csv")) 
LW_mod2 <- lm(log_mmax ~ log_lmax + class, fread("data/lw.csv"))

fig_s2 <- 
  fread("data/lw.csv") %>% 
  mutate(log_mmax_fit = predict(LW_mod2)) %>% 
  ggplot() +
  aes(log_lmax, log_mmax) +
  geom_point(size = 3) + 
  facet_wrap(.~class) +
  geom_line(aes(log_lmax, log_mmax_fit), col = mypal[1], size = 2) + 
  geom_abline(aes(intercept = LW_mod1$coefficients[1], slope = LW_mod1$coefficients[2]), lty=2, size=1.5, col="grey70") + 
  labs(x=expression(paste(log(L[infinity]), ", cm")), 
       y=expression(paste(log(M[infinity]), ", g"))) +
  theme_bw(base_size = 30) 

save_plot("output/figs/fig_s2.png", fig_s2, base_height = 10, dpi=300)
knitr::include_graphics("output/figs/fig_s2.png")

Figure S2: The relationship between body length and body mass for 85 invertebrate species. Two linear models are shown, the first with taxonomic class included as a predictor variable (solid red line) and the second without (dashed grey line).

Cut-off

fig_s3 <- 
  fread("data/cutoff.csv") %>% 
  mutate(group_ordered =  factor(group, 
                                 levels = c("Vertebrate",
                                            "Invertebrate"), ordered = TRUE)) %>% 
    ggplot(aes(log2(bin_mid), log_bin_density, fill = group_ordered)) +
  geom_vline(xintercept = 5, size = 2, lty = 2, col="grey70") +
  geom_point(size = 3,  pch = 21, key_glyph = draw_key_rect) + 
  geom_smooth(se = F,  size = 2, aes(colour = group_ordered), show.legend = F, method = "loess") +
  facet_wrap(~lat_region) + 
  annotation_logticks(base = 2) +
  xlab(expression(paste(log[2], "(M)"))) +
  ylab(expression(paste(log[2], "(N)"))) + 
  theme_bw(40) +
  scale_color_manual(values = c(mypal[2], mypal[1]), guide = "none") +
  scale_fill_manual(values = c(mypal[2], mypal[1]), name = "Community", labels = c("Fish-only", "Invertebrate-only")) +
  theme(
    axis.title.x = element_text(hjust = 0.45, vjust=0),
    legend.position = c(0.35, 0.85), 
    legend.background = element_rect(fill = "transparent"), 
    legend.key.size = unit(1, "cm"), 
    legend.key = element_rect(colour = "black", fill = NA),
    legend.title = element_text(size = 30),
    legend.text = element_text(size = 20))

save_plot("output/figs/fig_s3.png", fig_s3, base_height = 10, dpi=300)
knitr::include_graphics("output/figs/fig_s3.png")

Figure S3: Entire size spectra for invertebrate-only and fish-only reef communities. Dotted grey vertical line at \(2^5\) (= 32g) indicates the lower-bound size bin used as the cut-off. A smooth line is fitted using locally estimated scatterplot smoothing (LOESS).

Visiualisation

fig_s4 <- 
  fread("data/nass_fitted.csv") %>% 
    ggplot(aes(bin_mid, norm_density, colour = as.factor(site_id))) + 
    geom_point() +
    geom_line(aes(y = norm_density_fit)) +
    xlab("Mass bin, g") +
    ylab(expression(paste("Abundance, ", m^-2))) +
    scale_x_continuous(breaks = trans_breaks("log2", function(x) 2^x), trans = "log2", labels = trans_format("log2", math_format(2^.x))) +
    scale_y_continuous(breaks = trans_breaks("log2", function(x) 2^x), trans = "log2", labels = trans_format("log2", math_format(2^.x))) +
    facet_wrap(~ecoregion_trim) + 
    theme_bw(14) +
    theme(legend.position = "none",
          axis.title = element_text(size = 30), 
          axis.text = element_text(size = 14))

save_plot("output/figs/fig_s4.png", fig_s4, base_height = 10, dpi=300)
knitr::include_graphics("output/figs/fig_s4.png")

Figure S4: The normalized abundance size spectrum for 93 coastal ecoregions. Linear mixed effects models are fitted with site nested within ecoregion as random effects. Each site is individually coloured within the ecoregion. Axes are on the log-log scale.

Latitude zones

dodge <- position_dodge(width = 1)

fig_s5 <- 
  fread("data/lat_zone.csv") %>% 
  mutate(lat_zone_ordered = factor(lat_zone, levels = c("Tropical",
                                                        "Temperate",
                                                        "Polar"), ordered = TRUE)) %>% 
  mutate(name_ordered = factor(name, levels = c("slope_vert",
                                                "slope_comb"), ordered = TRUE)) %>% 
  ggplot(aes(lat_zone_ordered, value, fill = name_ordered)) +
  geom_hline(yintercept = -2, lty=2) +
  geom_violin(position = dodge) +
  geom_boxplot(width = 0.1, position = dodge, key_glyph = draw_key_rect) + 
  theme_bw(40) +
  labs(y = expression(lambda),
       x = element_blank()) +
  scale_fill_manual(values = c(mypal[2], mypal[1]), name = "Community", labels = c("Fish-only", "Fish and invertebrates")) +
  theme(
    axis.title.y = element_text(angle = 0, vjust = 0.5),
    axis.title.x = element_text(hjust = 0.45, vjust=0),
    legend.position = c(0.16, 0.9), 
    legend.background = element_rect(fill = "transparent"), 
    legend.key.size = unit(1, "cm"), 
    legend.key = element_rect(colour = "black", fill = NA),
    legend.title = element_text(size = 40),
    legend.text = element_text(size = 25)
  )

save_plot("output/figs/fig_s5.png", plot = fig_s5, base_height = 10, dpi = 300)
knitr::include_graphics("output/figs/fig_s5.png")

Figure S5: Reef size spectra slope estimates by latitudinal zone. Distribution of estimates of the site-level normalized abundance size spectrum slope (\(\lambda\)) for temperate and tropical sites. A dotted line at -2 is added to highlight the theoretical expected value.